So, I would like to know how much there is in difference between OGS and AGA ranks. There is a pretty big difference between OGS and Fox. I feel like 1d OGS is pretty equal to 5-6d on fox. I do feel like a 1d OGS is significantly stronger than 1d AGA as well. What about you guys?
Main topic. Now can be good time to collect some feedback.
Also playing on OGS, in the region from about 1k-3d, I’ve played people across that range, there really doesn’t seem to be too much of a strength difference whether I play a 1k or 3d.
Well first between a one level difference, it’s not a wall but a probability to win like lowering from 50 to 40%.
Second i like to compare graphs more as an instant ranking. A 1k evoluating between 4d-1k is surely not same strength as another between 1k-5k
OGS ranks may differ depending on whether you play mainly correspondence or live games. My impression is that correspondence ranks are 1 stone weaker than live ranks.
In addition people don’t always play seriously on the internet. Given two persons with the same internet rank, their IRL rank may differ by 4 stones.
WIth all those caveats in mind, my impression is that in the high SDK/low dan range, a player with EGF rank n has an OGS rank in the range [n, n+2]. It’s hard to find many examples though, most people I know don’t play a lot of ranked games on OGS or don’t have a reliable EGF rank.
I don’t know AGA ranks well enough to compare the OGS scale with the AGA scale.
My EGF rating is 3d and my OGS rating is 3d. Clossius (@Clossius1) has an OGS rating close to mine and his AGA rating is 4d. From that very small number of data points, it looks like mid dan OGS ratings (since the 2021 update) are about 1 rank tougher than AGA ratings.
I’m european, but i’ve played in few aga tournaments and i’ve gotten the impression that on sdk-level aga ranks are roughly 2 stones weaker than on OGS.
But of course that impression is based on very small sample size, and also both AGA and EGF ratings have suffered a lot because the lack of tournaments in the last 2 years.
The last rating system update OGS had was meant to put the OGS 1d rank somewhere between EGF 1d and AGA 1d, and then spread the ranks along the idea “rank difference = amount of handicaps for roughly 50/50* winrate”
*on average it was actually around 57/43 on whites advantage, but i guess thats the closest could be done with mathematical model that was used. Winning 4-6 games out of 10 seems close enough to be ~even.
I am EGF ranked not AGA, but it’s not the problem. OGS got more the problem, that the ranks are more random than elsewhere, cause of the small player base here. That is a matter, only solved with more member doing more ranked games.
I started playing more here, cause it’s no solution, to stay away, but I played 5ks stronger than some 2ks and so on. Thats why I still play on KGS more (and for some time Tygem),than here, even so, in the overall OGS is definitely the better side.
It’s a self fulfilling prophecy. People stay away, cause you can really often not be sure, how good your opponent really is, but cause people stay away, the ranks are more random.
I don’t understand why the number of players should affect the stability/validity of the ranks. Incidentally, I think KGS has fewer ranked games than OGS (but I don’t swear by it).
KGS has a system where the more you play, the more stable your rank becomes, whereas OGS always considers the last 15 games, maybe that’s it. I don’t know how Tygem works.
However, I am sorry that you have experienced this and that it bothers you so much that you prefer playing on other servers. I suggest you to think that rank is just a number, you are playing with people
Btw, I agree that ranks here are very unstable, without having to talk about extreme cases like me. But a 5k that is stronger than 2ks can almost surely become 1k+ by playing more.
Imho you can really make a comparison between online servers and AGA/EGF. I think the main reason, as already mentioned, is that many people are less engaged online and therefore perform worse, but occasionally they can play at their real level. And it’s okay, I think, I’m not talking about sandbagging.
I agree that OGS has a higher rating volatility than KGS, but I don’t think it has to do with the player base. AFAIK it’s more a design choice to select a lower or higher volatility for the rating system.
There are pros and cons to higher or lower rating volatility. You don’t like the higher volatility, but I also hear complaints that KGS ratings tend to get “stuck” at some point, causing players to create new accounts just to enable them to “rank up”.
Also, I suppose that a more volatile rating system contributes to reaching ratings that roughly fit new members’ levels more quickly.
I think a way trough the middle is better. But I think more people playing, would stabilize it at some point.
KGS2 had that issue, it’s resolved since the introduction of WHR.
Oh, I was unaware of the switch. When did that happen?
O I might have misremembered, I thought the KGS3 algorithm (2006-2007?) was changed to WHR when the study came out (2008), but can’t find a source to support that. WHR was tested on KGS data though, that much is evident.
The KGS2 algorithm was definitely changed when KGS3 (CGoban 3.x) came out and it bumped most SDK by about 4 ranks (I jumped from 6k to 2k), changes in the dan levels were a little less severe.
The 15-game sliding window was eliminated in Jan 2021 (anoek’s announcement).
I had read it here a few days ago, but it is actually out of date.
I would be surprised if this were the case. anoek investigated WHR and noticed the overhead was far too high. That’s why he chose glicko-2 instead.
I have not gotten the impression that KGS has significantly more server power than OGS?